Supercomputer Fugaku, who didn't aim for the world's best, ended up winning four world crowns.

*This article is based on Masakazu Kobayashi's Photo of Japan after Supercomputer Fugaku = iStock.com/gorodenkoff

Fugaku, the world's best domestically produced supercomputer born from reflection on Kyoto

Fugaku, which stands at the top of many supercomputers currently in operation in the world, is one of Japan's so-called national projects. born as a result. In recent years, the term "national policy" has not been used in a very good sense, such as "national policy investigation," but here we should understand it literally as "policy decided by the government." Or you could call it a national project.

In other words, we are required to contribute widely to the promotion of science and industry in our country as much as we invested a huge amount of national budget. This is the mission assigned to Fugaku.

Kyo, which was also developed by RIKEN/Fujitsu, casts a complicated shadow over this Fugaku as the previous national supercomputer.

In 2009, in the process of sorting out projects under the Democratic Party of Japan government at the time, Kyoto, which was still under development, was picked up as a target. With a total budget of more than 110 billion yen, a member of the House of Councilors Renho asked, "Isn't second place useless?"

There may be various points of view depending on the position placed, but at least it poses an essential problem that reconsiders the necessity of giant science such as "supercomputer" and "(nuclear / elementary particle) accelerator". There is no doubt that it was.

Kyo's bad reputation as "high performance but difficult to use"

In response to this, it seemed that the development plan for Kyoto would be temporarily frozen, but successive Japanese Nobel Prize winners were urgent reporters There was a fierce backlash, especially in the scientific community, where people held a press conference to express their concerns. The Hatoyama Cabinet at the time, seeing that public opinion had also leaned toward scientists, allowed the de facto budget to be reinstated, and continued the Kyoto development project.

Although it was not yet completed, Kyo, whose basic performance was measured, recorded a calculation speed of 8162 trillion flops (slightly short of the target of 1 quadrillion floating point operations per second), In the TOP500, it won first place for two consecutive terms in June and November 2011 (later, it also achieved the original target of 10 quadrillion calculations per second). In 2012, it was completed and started operation at the RIKEN Computational Science Research Center in Kobe.

*Editor's Note: When first published, the name of RIKEN Center for Computational Science was incorrect. Correct. (Added at 9:15 on April 5)

However, after the actual operation began, the K computer suffered from a reputation of being "high-performance but difficult to use." This was due in part to the fact that general application programs widely used in the industrial world could not be used on the K computer because the non-mainstream architecture (basic design) called "SPARC spark" was adopted at the time. It is the cause.